← Back to community benchmarks

gemma-4-26B-A4B-RotorQuant

M3 Pro (18c) · 36 GB · 4bit · 2026-05-20
Performance
16k
tokens
514.4
PP tok/s
12.3
TG tok/s
31851
TTFT (ms)
16.5
Peak mem (GB)
Hardware
Chip M3 Pro (18c)
Memory 36 GB
GPU Cores 18
Software
oMLX v0.3.9rc1
macOS macOS 26.4.1
Context 16,384
Performance by Context Length
Context PP tok/s TG tok/s Peak Mem
1k 523.2 41.4 14.3 GB view
4k 563.9 39.8 15.0 GB view
8k 553.1 33.6 15.3 GB view
16k 514.4 12.3 16.5 GB current
Batching Results
Batch Size TG tok/s Speedup
41.4 1.00×
58.9 1.42×
65.1 1.57×
82.9 2.00×